How exactly to install spaCy with GPU

[sdspieg]

I am trying to install spaCy on my wsl under Windows. I made an environment with python 3.8 and CUDA 11.3.
(spacy) root@Stephan-Office:~# nvcc --version nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2021 NVIDIA Corporation Built on Mon_May__3_19:15:13_PDT_2021 Cuda compilation tools, release 11.3, V11.3.109 Build cuda_11.3.r11.3/compiler.29920130_0
I did
pip install spacy[cuda113]
and then
pip install spacy-transformers
All of that seems to work fine
But when I then run this in python

import spacy

spacy.require_gpu()
nlp = spacy.load("en_core_web_trf")
print(nlp.pipe_names)  # Should include 'transformer'

I get this

Unexpected exception formatting exception. Falling back to standard exception
Traceback (most recent call last):
  File "/root/miniconda3/envs/spacy/lib/python3.8/site-packages/IPython/core/interactiveshell.py", line 3505, in run_code
    if async_:
  File "/tmp/ipykernel_267/487741322.py", line 4, in <module>
    nlp = spacy.load("en_core_web_trf")
  File "/root/miniconda3/envs/spacy/lib/python3.8/site-packages/spacy/__init__.py", line 51, in load
    return util.load_model(
  File "/root/miniconda3/envs/spacy/lib/python3.8/site-packages/spacy/util.py", line 465, in load_model
    return load_model_from_package(name, **kwargs)  # type: ignore[arg-type]
  File "/root/miniconda3/envs/spacy/lib/python3.8/site-packages/spacy/util.py", line 501, in load_model_from_package
    return cls.load(vocab=vocab, disable=disable, enable=enable, exclude=exclude, config=config)  # type: ignore[attr-defined]
  File "/root/miniconda3/envs/spacy/lib/python3.8/site-packages/en_core_web_trf/__init__.py", line 10, in load
    return load_model_from_init_py(__file__, **overrides)
  File "/root/miniconda3/envs/spacy/lib/python3.8/site-packages/spacy/util.py", line 682, in load_model_from_init_py
    return load_model_from_path(
  File "/root/miniconda3/envs/spacy/lib/python3.8/site-packages/spacy/util.py", line 539, in load_model_from_path
    nlp = load_model_from_config(
  File "/root/miniconda3/envs/spacy/lib/python3.8/site-packages/spacy/util.py", line 587, in load_model_from_config
    nlp = lang_cls.from_config(
  File "/root/miniconda3/envs/spacy/lib/python3.8/site-packages/spacy/language.py", line 1818, in from_config
    nlp = lang_cls(vocab=vocab, create_tokenizer=create_tokenizer, meta=meta)
  File "/root/miniconda3/envs/spacy/lib/python3.8/site-packages/spacy/language.py", line 185, in __init__
    util.registry._entry_point_factories.get_all()
  File "/root/miniconda3/envs/spacy/lib/python3.8/site-packages/catalogue/__init__.py", line 110, in get_all
    result.update(self.get_entry_points())
...
    colorsnormal, py3compat.cast_unicode(evalue_str))]
  File "/root/miniconda3/envs/spacy/lib/python3.8/site-packages/pygments/styles/__init__.py", line 92, in get_style_by_name
    raise ClassNotFound("Could not find style module %r" % mod +
pygments.util.ClassNotFound: Could not find style module 'default', though it should be builtin.

I'd very much appreciate it if somebody could post the exact versions of python, CUDA, etc. are required to make everything work properly.

[sdspieg]

I even tried downgrading to CUDA 10.1, but that needs gcc-7, which is not available in recent Ubuntu versions or in ppa:ubuntu-toolchain...

[sdspieg]

I already wasted 6 hours on this. Just a suggestion - couldn't you change the quickstart to avoid these kinds of problems? So that people can also be assisted with all the 'possibkle' choices, also for CUDA etc? And maybe also add wsl as an option on top under windows (as is the case for pytorch and rapids)?
Anyway I now tried this

- python -m venv spacy
- source spacy/bin/activate
- pip install -U pip setuptools wheel
- pip install -U 'spacy[cuda-autodetect,transformers,lookups]'
- python -m spacy download en_core_web_trf

And when I try this to check whether it works

import spacy

spacy.require_gpu()
nlp = spacy.load("en_core_web_trf")
print(nlp.pipe_names)  # Should include 'transformer'

I now get this

ValueError                                Traceback (most recent call last)

Cell In[1], line 1
----> 1 import spacy
      3 spacy.require_gpu()
      4 nlp = spacy.load("en_core_web_trf")

File [~/miniconda3/envs/spacy/lib/python3.11/site-packages/spacy/__init__.py:14](https://file+.vscode-resource.vscode-cdn.net/d%3A/Google%20Drive/RuBase/EUCOM/Kremlin%20watchers/Bibliometrics/~/miniconda3/envs/spacy/lib/python3.11/site-packages/spacy/__init__.py:14)
     11 from thinc.api import prefer_gpu, require_gpu, require_cpu  # noqa: F401
     12 from thinc.api import Config
---> 14 from . import pipeline  # noqa: F401
     15 from .cli.info import info  # noqa: F401
     16 from .glossary import explain  # noqa: F401

File [~/miniconda3/envs/spacy/lib/python3.11/site-packages/spacy/pipeline/__init__.py:1](https://file+.vscode-resource.vscode-cdn.net/d%3A/Google%20Drive/RuBase/EUCOM/Kremlin%20watchers/Bibliometrics/~/miniconda3/envs/spacy/lib/python3.11/site-packages/spacy/pipeline/__init__.py:1)
----> 1 from .attributeruler import AttributeRuler
      2 from .dep_parser import DependencyParser
      3 from .edit_tree_lemmatizer import EditTreeLemmatizer

File [~/miniconda3/envs/spacy/lib/python3.11/site-packages/spacy/pipeline/attributeruler.py:6](https://file+.vscode-resource.vscode-cdn.net/d%3A/Google%20Drive/RuBase/EUCOM/Kremlin%20watchers/Bibliometrics/~/miniconda3/envs/spacy/lib/python3.11/site-packages/spacy/pipeline/attributeruler.py:6)
      3 import srsly
      4 from pathlib import Path
----> 6 from .pipe import Pipe
      7 from ..errors import Errors
      8 from ..training import Example
...
   2724 if is_keyword or not name.isidentifier():
-> 2725     raise ValueError('{!r} is not a valid parameter name'.format(name))
   2727 self._name = name

ValueError: 'in' is not a valid parameter name

[sdspieg]

Just trying to help others with the same problem - here's (succinctly put) what worked for me. (And please note - I'm doing this on a Windows machine, running wsl2)

• CUDA 11.5
• Python 3.9.17
• python -m venv spacy
• source spacy/bin/activate
• pip install -U pip setuptools wheel
• pip install -U 'spacy[cuda-autodetect,transformers,lookups]'
• python -m spacy download en_core_web_trf

[svlandeg]

Hi @sdspieg, I'm sorry to hear that it was so difficult to get the installation up and running. Thanks for sharing the solution for your system!